Micro Programmable Object Navigation Gadget (µ-PONG) for Studying Electroosmotic Flow in a PDMS Microchannel Brittany Rohrman Background and Significance: One of the most useful features of microfluidic devices, in the context of biological research, is their ability to serve as tiny fluidic enclosures for either individual or small collections of cells. The small volume of such enclosures can enable the in vitro study of cellcell interactions as well as paracrine and autocrine signaling events that would be difficult to achieve in traditional large Petri dish cell culture environments, which typically have extremely large (and non-physiological) ratios of cell media fluid volume to cell volume. Microfluidic devices are more comparable to in vivo tissue environments because they are fabricated to house cells in small fluidic volumes. However, one constraint on the long-term observation of cells enclosed in tiny fluidic enclosures is the need to replenish the nutrient solutions surrounding the cells. Another constraint when utilizing microfluidic devices is the need for very low flow rates capable of manipulating cells without disrupting pre-established biochemical diffusion gradients. These low-flow techniques typically require expensive (~$3000) precision syringe pumps to provide on-demand, extremely low fluid flow rates. An attractive alternative to this expensive mechanical syringe pump technology is an electro-kinetic technique that can induce low volume electroosmotic flow in microchannels. Objective: The goal of my senior honors thesis project is to create an automated testing system (µPONG) for characterizing electroosmotic flow (EOF) in poly(dimethylsiloxane) (PDMS) devices. My hope is that this system will be used to characterize EOF in particular PDMS devices intended for biological research in order to demonstrate the possibility of utilizing EOF as a pumping system in such devices. The basic premise of my research is that by using the computer-controlled µ-PONG system to control the movement of a polystyrene bead in a PDMS microchannel, the recorded motion of the bead will serve as a measure of the electroosmotic flow strength over time. Hence, this characterization may be an important factor in the design of biological experiments. Once the system is optimized for well-controlled fluid delivery, the µPONG system may also serve as a tool for manipulating particles or cells in PDMS microfluidic devices. Introduction: Poly(dimethylsiloxane), or PDMS, is a polymer often used to construct microscopic or nano-scale channels. Because PDMS is permeable to gases, nontoxic to cells, and optically clear,1 these channels often serve as an optimal environment in which electroosmotic flow (EOF) is used to manipulate cells2 and biomolecules3 for study. In experiments utilizing EOF, a potential difference is applied across an oxidized PDMS channel filled with a buffer. Cations from the buffer are attracted to anions bound to the walls of the channel, forming what is known as the electric double layer (EDL), which is only a few nanometers wide (Figure 1). 2 Oxidized PDMS EDL Electroosmotic Flow Figure 1: Electroosmotic flow (EOF). When an electric field is applied across the channel, positive ions from electric double layer (EDL) move toward the negative electrode, dragging the rest of the solution in that direction. The potential difference causes the layer of cations to move toward the negative electrode, pulling the rest of the solution in the same direction via viscous drag.4 This phenomenon is known as electroosmotic flow. The flow is often quantified by the electroosmotic mobility, eo , which is defined by (the permittivity of the solution), (the zeta potential), and (the viscosity of the solution).5 Another common measure of electroosmotic flow is the electroosmotic velocity, veo, which is closely related to the electroosmotic mobility. The electroosmotic velocity is defined as veo E eo , where E denotes the applied electric field.6 Electroosmotic flow has been implemented in a wide variety of applications, often serving as a pump in lab-on-a-chip (LOC) microfluidic devices.7 LOC devices allow bioassays to be performed on a much smaller scale, reducing the time and volume of reagents required for traditional methods. Electroosmotic flow is ideal for use in LOC devices because it requires no moving parts and may be controlled very precisely, unlike pressure-driven flow. Another potential advantage of EOF is that the velocity of the fluid is the same throughout the cross section of the channel (plug flow), while the velocity profile of pressure-driven fluid flow is parabolic (Poiseuille flow). However, the strength of the EOF in oxidized PDMS channels has been reported to attenuate over a period of just a few hours.8 This research aims to quantify this effect by studying the movement of fluorescent polystyrene beads manipulated by EOF in a PDMS microchannel. Once these time-dependent characteristics of electroosmotic flow are known, use of the flow may be optimized for various microfluidic applications. 3 Methods: Microfabrication of PDMS devices PDMS microchannels were made using standard microfabrication and soft lithography methods (Figure 2). First, SU-8 photoresist on a silicon support is exposed to ultraviolet light through a mask. The light cures the exposed photoresist, leaving the photoresist obscured by the mask uncured. The uncured photoresist is then dissolved, leaving a raised pattern defined by the mask called a master. After further surface treatments, uncured (liquid) PDMS is poured onto the master and degassed to remove bubbles. The PDMS is then baked overnight to solidify the polymer. After exposure to an oxygen plasma for 30 seconds to create negative charges (Si-0–) on the PDMS surface (for our devices, a PDC-32G Plasma Cleaner/Sterilizer was used on the “high” power setting), the PDMS is irreversibly bonded to a clean glass slide. Our devices were filled with 10 mM monobasic potassium phosphate buffer at pH 7.0 immediately after plasma bonding. Carboxylated or uncarboxylated fluorescent polystyrene beads (Polysciences, Inc. Fluoresbrite YG microspheres, CV = 5%) with diameters of 3 or 10 microns were used. Several microliters of beads at concentrations between 0.01 and 0.001% solid were deposited into one end of the channel using a micropipette. In some experiments, the pressure was equalized by pumping additional buffer into the channel. Then a computer program was used to control and/or record the movement of the bead. UV light mask 1 photoresist and silicon wafer 2 master 3 PDMS poured into master 4 PDMS bonded to glass Figure 2: Soft lithography and microfabrication of PDMS channels. The µ-PONG system The µ-PONG system consists of a DAQ board, an external voltage switching circuit, and a camera, which are controlled by a computer program written in a LabView programming environment (Figure 3). The computer is connected to a National Instruments USB-6009 device capable of producing an analog voltage ranging from zero to five volts or a digital control signal (on = 0 volts, off = 5 volts) (Figure 4). To increase the range of possible voltages used for experiments, an external circuit was designed and built to amplify the USB analog voltage by a factor of 1.83 (Figure 5). A digital voltage channel from the USB device is also connected to the 4 circuit in order to control the directional sense (or sign) of the voltage at the output leads, such that changing the state of the digital signal reverses the direction of the potential difference across the leads. The analog output signals from this external circuit establish a voltage between the electrodes that are inserted at each end of the microfluidic channel (Figure 6). Thus, a voltage is set up across the channel that causes electroosmotic flow. Gold electrode Goal posts PDMS Glass Polystyrene bead -fluidic device Control box Objective -scope stage Dichroic mirror Excitation Ex filter Em filter Camera / detector Computer Figure 3: Schematic diagram of the µ-PONG system. Two electrodes from the control box are inserted at each end of the PDMS microchannel, which is imaged through a camera via fluorescence microscopy. A LabView program displays the image from the camera and communicates with a USB device and external circuit housed inside a control box. When the bead reaches a virtual goalpost, the program detects the bead and reverses the direction of the voltage across the channel, keeping the bead moving between the goalposts. The LabView program continually retrieves and displays an image of the microchannel in real time. Based on the image, the user utilizes the software to specify two virtual “goalpost” regions of interest located at opposite ends of the imaged channel, typically about 200 microns apart. The pixel intensities of the goalposts are compared to a user-specified target intensity value in order to determine if a bright bead has entered a goalpost region. When that target intensity value is reached or exceeded, the program sends a signal via the USB port to the DAQ device, changing the state of the digital voltage channel connected to the external circuit. That signal reverses the polarity of the voltage across the PDMS microchannel, thereby reversing the direction of the EOF (see Appendix A for block diagram of LabView program). This strategy works well because fluorescence microscopy is used to capture the image of the channel, causing the bead to appear brighter than anything else in the channel. Thus, whenever the bead reaches a region of interest, the reversal of the EOF causes it to move toward the other region of interest, where the process is repeated. The bead moves back and forth between the regions of interest much like a ball moves back and forth during a game of ping pong, allowing experiments to run for long periods of time. 5 Figure 4: The NI-USB 6009 device relays signals from the computer to the circuit. Figure 5: The external circuit amplifies the signal from the USB device and changes the direction of the voltage. Figure 6: The electrodes at each end of the PDMS channel cause EOF, which moves the bead. The LabView user interface Before the program is run, the user can specify channel width, device age, and other comments in a table on the front panel labeled “parameters” (Figure 7). On the front panel, the user may also specify the physical digital and analog output channels of the USB device as well as their minimum and maximum voltages, although these are already set to default values. The user clicks the “run” arrow to execute the program. Instructions for saving the parameters and subsequent data are displayed before the user is asked to specify a *.lvm filename and location to save the parameter table. On the front panel, the user specifies the desired analog output voltage to send to the USB device via a slider. The user must also specify the pixel locations of the regions of interest. The “left indicator” and “right indicator” are illuminated each time the bead is detected in the specified region of interest during the experiment. Figure 7: LabView interface. The user may specify the location of the virtual goalposts, the detection threshold, and the voltage 6 across the channel. The user may also pause the EOF, save data, record a video, and specify the mode of the system (“manual” or “automatic”). The user may run the program in a manual or automatic mode by clicking the “Manual/Auto” button on the front panel. The “manual” mode allows the user to switch the polarity, and thus the direction of the EOF, by clicking the “Left/Right” button on the front panel. The “auto” mode detects the bead at the regions of interest and reverses the polarity automatically. The user may also turn on or off the voltage across the channel by clicking the “Pause/Run” button on the front channel. The “pause” mode simply sets the voltage to zero, while the “run” mode establishes the voltage set on the slider. To record the analog output voltage, the voltage polarity, and whether the desired analog output voltage is “running” or “paused,” the user must click the “Record” button on the front panel. When the record button is on, these parameters as well as the timestamp are recorded each time the voltage polarity changes. When prompted, the user should select the file where the parameters were saved previously in order to append the data to the same file. The user may also save a *.avi movie by clicking the “Record” button located below the image display. When the user clicks the “stop” button on the front panel, the program ceases to execute. Voltage-switching circuit One of the design goals of the µ-Pong system was to incorporate an easy way to vary the magnitude of the applied voltage delivered to the microfluidic device. For this purpose, we utilized a computer-controlled 12-bit digital-to-analog converter, which is a component of the National Instruments USB-6009 device. However, the output of this particular device is limited to a unipolar 0-5V range, lacking the ability to reverse the applied voltage in order to reverse the motion of the flow. Because reversing the direction of the EOF is the main function of the µ-PONG system, a circuit utilizing integrated circuit analog switches (74HC4066N) was designed to provide a variable magnitude and reversible voltage system. The circuit is powered by a power supply that converts 120V AC voltage to 15V and -15V (DC voltage). A positive regulator converts the 15V voltage to 5V and a negative regulator converts the -15V voltage to -5V in order to power five LEDs that serve as indicators and the integrated circuits (a hex inverter, operational amplifiers, and a digital switch - see Appendix B for the analog switch datasheet). One LED is connected to ground such that when it is powered, a voltage is established across it, indicating that the power is on (see Figure 8). Digital signals two and three are also connected to current-limited LEDs to indicate whether the system is in manual or automatic mode and whether the system is paused or running. The voltage-switching function of the µ-PONG circuitry works by connecting output amplifiers to either a signal which is equal to the analog input voltage or a signal which is equal to minus the analog input voltage, depending on whether leftward motion or rightward motion is required by the control logic signals. Digital signal one serves as the switch for changing the polarity of the voltage across the microfluidic channel. When the input signal is at logic 1, “TTL high,” the LED indicating leftward motion is illuminated; when the signal is at logic 0, “TTL low,” the LED indicating rightward motion is illuminated, specifying the direction of the voltage across the channel (and the direction of the EOF). The signal is split between two branches, the first containing one inverter and the other containing two inverters. Thus, the control signals at the two branches are 7 always at opposite states: when one is at logic “1,” the other is at logic “0.” Because the particular analog switch integrated circuit utilized in this design requires an input control signal which defines logic 1 and logic 0 at levels which are not compatible to standard TTL logic levels, level translating circuitry consisting of diodes and resistors was utilized to convert TTL level signals to the logic ranges required by the analog switch circuit. Figure 8: Schematic diagram of the external voltage-switching circuit. Five LED’s serve as indicators of the power, system modes, and direction of the EOF. Digital signal one serves as the polarity switch by routing the analog input voltage signal through one of two switches that cause the signal to be multiplied by either one or negative one. The output signals of two operational amplifiers serve as the voltages sources for the electrodes that establish the electric field across the channel. Based on the states of the logic switches, the analog input voltage will either pass through an operational amplifier, multiplying its value by negative one, or remain unchanged. (Because digital signal one controls this function, that voltage channel serves as the voltage polarity switch.) The analog voltage then is amplified by a factor of 1.83 through one operational amplifier and then sequentially multiplied by negative one through another. Thus, the output electrode connector signals are at +1.83 (Vin) and 1.83 (Vin). Wires from these operational amplifier outputs are connected to gold electrodes at each end of the channel. Hence, at one electrode, the voltage is +V and at the other, the voltage is –V. If digital signal one (the polarity switch) changes state, the input voltages are then multiplied by negative one, changing the direction of the EOF in the channel. The voltage switching circuit, DAQ board, and power supply are housed in a metal box (see Figure 9). The LEDs are visible from the outside of the box, serving as indicators of whether the power is on, whether the mode is “automatic” or “manual,” whether the system is “paused” or “running,” and whether the direction of the EOF is 8 “right” or “left.” Two binding post jacks on the outside are the sources of the voltages to be established at each end of the channel. Figure 9: Control box containing the USB DAQ device and external circuit. LED indicators show the status of the system, and two binding post jacks serve as the sources of the voltage to be established at each end of the channel. Three-dimensional particle tracking Due to pressure differences at each end of the channel, the fluid flow in our PDMS microchannels may consist of both pressure-driven and electroosmotic flow. We expect that the speed distribution of particles in the z-dimension of a channel should be quite different for the two flow types because the driving force for EOF is generated at the walls of the channel, whereas in the case of pressure-driven flow, the channel walls serve as friction boundaries that limit the flow velocity. Thus, in order to fully characterize the fluid flow profile in our PDMS microchannels, the velocity of the beads in three dimensions must be known. By capturing videos with a camera attached to a microscope, the velocity of beads in the focal plane may easily be measured. However, because the beads near the bottom of the channel would require one focus setting, and beads near the top of the channel would require a completely different focus setting, the three-dimensional velocity profile of the beads requires a way to characterize the zposition of the bead. One possible way to characterize the z-dependence of particle velocity would be to collect data at many different z-values and measure the velocity of in-focus beads at each altitude. As an alternative technique, a z-axis microscope servo system could be developed in order to automatically bring a bead into focus, track the bead across the field of view, and record its z-value over time. The technique which I have investigated as part of this research involves another methodology, in which the z-axis position of the bead used for our velocity measurements is determined by the information encoded in the out-of-focus blur. The zcoordinate of a point source may be found by taking advantage of its point spread function (PSF), which is defined as the response of an imaging system to a point source. Because small fluorescent polystyrene beads are quite uniform in size (2.986 0.083 um), they are small enough to behave approximately like point sources at the magnifications used in these experiments. A point source radiates a complete spherical wave such that the placement of the lens intercepts a slice of the sphere (see Figure 10). The radius of the image becomes larger as the defocus increases. For even larger displacements, the image appears as a diffraction pattern of rings, the outermost of 9 which is the brightest ring.9 Thus, the PSF of the bead may be used to identify the distance of the bead from the focal plane. Retrieved with permission from: K. Seale, C. Janetopoulos, and J. Wikswo, ACS Nano 3, 493 (2009). Figure 10: The 3-dimensional point-spead function (PSF) of a point source. The collection cone of the microscope objective, shown here as the “lens,” intercepts a two-dimensional slice of the PSF, which appears as a pattern of interference rings. Several strategies for using defocused images to determine the z-coordinate of the particle position have been developed, including matching the radius vector of a reference image to that of an experimental image;10 measuring the diameter of the outermost ring of a defocused particle and using a mathematical formula to determine the z-coordinate;11 and creating and using a calibration curve based on the radius of the outermost ring of several reference images.12 The approach used for our data is similar to the lattermost strategy. Each experimental image was compared to several reference images with known z-coordinates. The z-component of the velocity may be computed by determining which reference image matched best and performing a calculation of the average velocity over several frames. Initially I used convolution filters to match the reference images to the experimental image. However, the shape of the intensity distribution prevented an accurate match, even when normalized several different ways. Better results were obtained by using a least squares algorithm. For example, if a reference image is represented by a matrix with entries rij and the experimental image is represented by a matrix with entries xij , the reference and experimental image match best when S is minimized: S rij xij . 2 j 1 i 1 10 To verify that the least squares matching algorithm produced accurate results, S was computed for several experimental images with known z-coordinates. When the reference image and experimental image corresponded to the same z-coordinate, the S value was about an order of magnitude less than the S values for all the other images. See Table 1 below for an example in which a five-tick image was compared with four reference images. (The microscope used in these experiments (Zeiss Axiovert 25) has a z-axis focusing control knob which incorporates uniformly spaced markings, or “ticks,” which can be used as a quantitative measure of the microscope stage z-axis position. Each “tick” corresponds to 9.6 microns above the focal plane.) Note that in Table 1, when the z-coordinates were the same for the experimental and reference image (five ticks), S was approximately one order of magnitude smaller than any other S value. Least squares test: 5-tick experimental image z of reference image S 1 tick 237856 3 ticks 472564 5 ticks 46473 7ticks 445519 Table 1 I initially made reference images by finding the maximum pixel value (which was usually located approximately in the center of the off-focus image) and cropping a 71-by71-pixel region of interest (ROI) around that pixel. However, when using this method, the maximum pixel of the experimental images was often located very far from the center of the image. To avoid this problem, I instead adopted a centroid-finding method to determine the center of the image. The calculation shown below, in which I is intensity, is analogous to finding the center of mass of an object: xCM , yCM xij I ij I ij , y I I ij ij ij . Each centered 71-by-71 reference image was then normalized by multiplying each pixel by 100 and dividing by the maximum value, creating an image with a maximum value of 100. Five or six reference images for each tick were averaged together to create a final generalized reference image. This was done for z-coordinates of 0-5 ticks at half-tick intervals (Figure 11). 11 Figure 11: Reference images and their intensity profiles used for determining the z-value of beads imaged during our experiments. Each tick corresponds to 4.8 microns in the vertical direction. The matching algorithm was written as a macro for ImageJ, a Java-based image processing package (see Appendix C). (All images must be converted to 32-bit TIFF format prior to running the algorithm in order to avoid pixel saturation.) When the macro is run, the user is prompted to specify the directory where the image sequence of experimental data is located. The macro also prompts the user to choose the directory where the reference images are located. After the first experimental image in the sequence is opened, the user is prompted to draw a small rectangle around the bead of interest. The macro then crops the first image down to a 71-by-71-pixel ROI centered on the centroid of the image. The location of this ROI is saved in a new folder called “roi” in the same directory. These files may be opened later to ensure that the program has accurately tracked the particle. Like the reference images, each experimental image is normalized to have a maximum value of 100. Next, each reference image in the userspecified directory is opened, and S is computed for each reference image. The filename of the reference image for which S is a minimum is displayed in a “log” window along with the x- and y-coordinates of the centroid. The algorithm repeats the entire process for each experimental image in the specified directory, displaying the location of the bead for each video frame. Using these x-, y-, and z-coordinates and the frame rate of the video capturing software, an approximation of the particle velocity in threedimensions may then be computed. Results and Discussion: Initial experiments confirmed that the µ-PONG system successfully detected the bead at each goalpost and reversed the EOF upon detection (Figure 12), allowing experiments to run for extended periods of time. We predicted that the time the bead 12 required to travel from one goalpost to the other (the “transit time”) would remain fairly constant over short time scales, since the attenuation of EOF should occur over several hours.8 Since the same volume of fluid should be displaced for every transit, we also expected that hydrodynamic flow due to differences in fluid height at the ends of the channel would cancel out in each direction, causing no net fluid displacement. However, our results indicated that the flow rate varies over periods of tens of seconds (Figure 13a). Figure 12: A bead controlled by the µ-PONG system moving between two goalposts. Transit Time vs. Time 35 Transit Time (s) 30 25 20 15 10 5 0 0 50 100 150 200 250 300 350 400 Time (s) Figure 13a: The time required for the bead to travel from one goalpost to another for each transit, plotted versus time. 13 Transit time in each direction vs. time 35 30 Right Transit Time (s) Left 25 20 15 10 5 0 0 50 100 150 200 250 300 350 400 Time (s) Figure 13b: The time required for the bead to travel from one goalpost to another for each transit, plotted separately for each direction versus time. After deconstructing the time required for the bead to travel between goalposts in each direction (Figure 13b), it appeared that the speed of the bead decreased when traveling to the right but increased over time when moving leftward. The bead then approached its original speed in both directions after about 350 seconds. Further, the decrease in speed rightward was about fifteen seconds per transit at its slowest point, but the increase in speed leftward was always less than five seconds. We surmised that since the initial transit times to the left and right were not equal, a net hydrodynamic flow was present, causing the superposition of pressure-driven flow and EOF to complicate the experiment (Figure 14). Thus, we used a syringe pump to equalize the pressure before the start of our next trial. In the absence of EOF, the pump introduced more buffer into the channel until the flow of beads stopped, indicating that the pressure was equalized. However, the results indicated that although the pressure imbalance was corrected, the transit times still varied widely throughout the experiment, much like the first experiment (Figure 15). PDMS Hydrodynamic pressure Electrode Glass slide _ Feof Fh Bead + Goalpost Figure 14: Superposition of pressure-driven and electroosmotic flow. If the height of the fluid at one end of the channel is different from the height at the other end, the imbalance causes pressuredriven flow. 14 Transit Time vs. Time 16 Transit Time (s) 14 12 10 8 6 4 2 0 0 50 100 150 200 Time (s) Figure 15: The time required for the bead to travel from one goalpost to another for each transit, plotted versus time. For this experiment, additional buffer was introduced into one end of the channel to eliminate the pressure imbalance in the channel. In all experiments, the bead appeared to move in all three dimensions, instead of only in the direction parallel to the channel walls. Initially, the ImageJ plugin, SpotTracker, was used to determine the trajectory of the bead in the channel. However, this program required the user to click on the bead in each frame in order to compute the velocity and contained no z-axis information, although the bead drifted out of focus, indicating motion along the z-axis. Thus, the three-dimensional particle tracking ImageJ macro was developed to include this information as well as automate the determination of the x- and y-coordinate of the bead. A reason that the bead may not only be moving along the x-axis is that, as mentioned above, a pressure imbalance is present. Because pressure-driven flow has a parabolic velocity flow profile while EOF has a flat profile (Figure 16), the bead may be spinning, causing it to move in the direction of the z- and y-axis. The beads also adhered to the channel walls after a period of about thirty minutes, indicating that electrostatic or covalent bonds may form between the beads and the wall, perhaps because the beads accumulate charge. As a test, three-micron carboxylated polystyrene beads were compared with standard three-micron polystyrene beads. Many more of the uncoated beads adhered to the channel walls than the carboxylated beads, which adhered to the walls only after a substantially longer time, if at all. By comparison, ten-micron polystyrene beads adhered to the walls almost immediately, suggesting that the weight of the beads may cause sinking and adhesion. These data suggest that an accurate characterization of the velocity profile in three dimensions is absolutely essential. 15 Electroosmotic flow Hydrodynamic flow Parabolic profile Flat profile Figure 16: Hydrodynamic flow has a parabolic profile, while electroosmotic flow has a flat profile. In order to test the accuracy of the reference-matching ImageJ macro, we decided to focus on measuring pressure-driven flow because its properties have been well studied. Indeed, an analytical solution exists describing the velocity profile of the fluid within a rectangular channel of a given height (height = 2b) and width (width = 2a). In the equation of the velocity profile u(y,z), u ( y, z ) cosh(i z / 2a) cos(i y / 2a) 16a 2 P , (1)(i 1) / 2 1 3 L i 1,3,5,... i3 cosh(i b / 2a) L is the length of the channel (0.02 meters), µ is the viscosity of water (0.00089 N/m2), and P is the pressure difference across the channel. The measured velocities were used to find a numerical value of P in order to calculate the theoretical profile. Figure 17: The theoretical velocity profile is shown as a green mesh, while the experimental data points are shown in blue. The 16 blue line is the residual between the velocity profile and experimental data. Residual (m/s) Initially, the experimental data were analyzed by plotting the velocity of the beads at each z-value (see Appendix D). Parabolas that were fit to these data demonstrate that the velocity flow profile of the fluid flow in the channel is parabolic except at very low z-values, due to the effects of friction at the surface of the channel. A plot of y- and zcoordinates versus velocity also demonstrated that higher velocities generally occur near the middle of the channel, while the flow is relatively slow at the channel edges. However, in order to fully characterize the velocity profile in three dimensions, the theoretical velocity curve was calculated and plotted with the experimental data of the bead velocities analyzed in the channel (see Figure 17). The measured data seemed to be consistent with the calculated profile expect for low z-values, due to friction near the edges. The beads at that height probably rolled along the bottom instead of moving with the flow of the fluid. Beads with z-values greater than about twenty microns were not observed, perhaps because the density of the beads was lower than that of water, causing the beads to sink. Differences between the calculated profile and the experimental data were near or below 0.0001 m/s (see Figure 18). These residuals may be attributed to approximating the z-value to 4.8 micron intervals. However, these data suggest that the reference-matching algorithm worked fairly well in determining the velocity profile in three dimensions and may therefore be used to characterize EOF. Y (meters) Figure 18: The difference between the calculated velocity and the velocity determined experimentally for each y-value. To demonstrate that beads may be found at all positions in the channel, the motion of beads flowing under hydrostatic pressure-driven force was recorded and analyzed in an experiment in which salt water was used as a buffer. Because the density of polystyrene is 1.05 g/mL, a solution of 10% NaCl in water was made to match the density of the polystyrene. As shown below (Figure 19), beads were found at all heights throughout the channel, although most beads were found near the center. This was due to the fact that the beads move faster in the center of the channel, thereby allowing more beads to be observed there during a given time. Hence, experiments may 17 be done with a buffer of about the same density as polystyrene so that the velocity profile may be calculated based on beads at all z-values in the channel. Bead Height Distribution 40 Number of Beads 35 30 25 20 15 10 5 0 0 4.8 9.6 14.4 19.2 24 28.8 33.6 38.4 43.2 48 Z (microns) Figure 19: The number of beads observed at each z-value. Conclusions and Future Directions: The results of our experiments have shown that the µ-PONG system precisely controls the movement of polystyrene beads by manipulating the electroosmotic flow within a PDMS microchannel. The analysis software, an ImageJ macro, has proven to be approximately accurate in determining the three-dimensional velocity profile of pressure-driven flow in the channel. Thus, the system and analysis methods should prove useful for characterizing EOF in the channel over time. In the future, this system should be used to study the three-dimensional velocity profile of EOF in the channel to characterize its attenuation over time. We expect that the analysis of such experiments would demonstrate that pressure-driven flow and EOF occur simultaneously. Further studies would also determine the reason that the motion of the beads fluctuates as observed, implicating the properties of the flow. Once the EOF is characterized for a single channel, the system may then be used to study the fluid dynamics through more complicated microfluidic devices. This information may then be used for designing and optimizing those devices as well as predicting the movement of fluid within the channels. This system may also be used to deliver fluid in a tightly controlled manner in a microfluidic channel. Such applications may include LOC devices that require careful control of a fluid pump. Other applications include the manipulation of micro- or nanoscale objects, such as particles or cells. Hence, µ-PONG should serve as a method for discovering the properties of fluid flow in microdevices and may potentially serve as a part of other microfluidic systems for studying biology. 18 References: 1 2 3 4 5 6 7 8 9 10 11 12 S. K. Sia and G. M. Whitesides, Electrophoresis 24, 3563 (2003). L. Cui, D. Holmes, and H. Morgan, Electrophoresis 22, 3893 (2001). A. E. Cohen and W. E. Moerner, Physical Review Letters 98, 116001 (2007). D. A. Skoog, F. J. Holler, and S. R. Crouch, Principles of Instrumental Analysis (Thomson Brooks/Cole, Belmont, CA, 2007). W. Hellmich, J. Regtmeier, T. T. Duong, R. Ros, D. Anselmetti, and A. Rox, Langmuir 21, 7551 (2005). G.-B. Lee, C.-H. Lin, K.-H. Lee, and Y.-F. Lin, Electrophoresis 26, 4616 (2005). X. Y. Wang, C. Cheng, S. L. Wang, and S. R. Liu, Microfluidics and Nanofluidics 6, 145 (2009). X. Ren, M. Bachman, C. Sims, and G. P. Li, Journal of Chromatography B 762, 117 (2001). M. Mansuripur, Journal of the Optical Society of America A - Optics Image Science and Vision 3, 2086 (1986). Z. Zhang and C. Menq, Applied Optics 47, 2361 (2008). M. Speidel, A. Jonas, and E. Florin, Optics Letters 28, 69 (2003). S. Peterson, H. Chuang, and S. Wereley, Measurement Science & Technology 19 (2008). 19 Acknowledgements: I would like to gratefully acknowledge Phil Samson, Dmitry Markov, and John Wikswo for their mentorship and assistance throughout this project. I would also like to thank Dawit Jowhar for his help in creating the LabView program by designing the imaging and detection components of the system. Finally, I would like to acknowledge the Vanderbilt Institute for Integrative Biosystems Research and Education, the William A. and Nancy F. McMinn Foundation, and the Systems Biology and Bioengineering Undergraduate Research Experience for the opportunity and funding to do this research. 20 Appendix A: LabView Block Diagram NO YES Switch polarity? AND Start session MANUAL Manual/auto mode? AUTO Create buffer NO Run? Find connected cameras? NO YES User Inputs Control camera settings YES Set min/max voltage Send zero volts to channel Acquire image Display image YES Send analog voltage to channel Paused? NO Set digital channel Acquire intensity of goalpost pixels Is pixel above threshold? NO (Left goalpost) Set pixel threshold YES Was threshold previously reached at right goalpost? Set analog channel voltage Save data? NO NO YES 21 YES Change state of digital channel one Did state of channel change from previous state? NO YES Write data to file Is pixel above threshold? (Right goalpost) AND NO YES Was threshold previously reached at left goalpost? NO YES Appendix B: Analog Switch Datasheet 22 23 24 Appendix C: 3-D Tracking ImageJ Macro dirSeq = getDirectory("Choose the directory where the image sequence is located"); files = getFileList(dirSeq); File.makeDirectory(dirSeq+"\\roi") dir = getDirectory("Choose the directory where the reference images are located"); list = getFileList(dir); for (i=0; i<files.length; i++) { open(dirSeq + files[i]); currentFile = File.getName(dirSeq + files[i]); if(i==0){ getSelectionBounds(p,q,w,h); if(p==0){ waitForUser("Draw a small rectangle around the bead, then click OK"); getSelectionBounds(p,q,w,h); xcenter = round(p + w/2 - 35); ycenter = round(q + h/2 - 35); } } run("Specify ROI", "roi width=70 height=70 x="+xcenter +" y="+ycenter+" slice=1"); run("Crop"); showStatus("Finding centroid..."); width = getWidth(); height = getHeight(); yadd = 0; xadd = 0; total = 0; for (y=0; y<height; y++) { if (y%20==0) showProgress(y, height); for (x=0; x<width; x++) { intensity = getPixel(x,y); xnum = intensity*x; ynum = intensity*y; yadd = yadd + ynum; xadd = xadd + xnum; total = total + intensity; } } xcentroid = xadd/total; ycentroid = yadd/total; xcorner = round(xcentroid - 25); ycorner = round(ycentroid - 25); showStatus("Cropping region of interest..."); run("Specify ROI", "roi width=51 height=51 x="+xcorner+" y="+ycorner+" slice=1"); saveAs("Selection", dirSeq+"\\roi\\"+currentFile+".roi"); run("Crop"); 25 showStatus("Normalizing..."); showStatus("Finding pixel with largest value..."); width = getWidth(); height = getHeight(); max=0; for (y=0; y<height; y++) { if (y%20==0) showProgress(y, height); for (x=0; x<width; x++) { value = getPixel(x,y); if (value>max) { max = value; xmax = x; ymax = y; } } } run("Divide...", "value="+max); run("Multiply...", "value=100"); large = 1000000000; tick = "unknown"; for (j=0; j<list.length; j++) { open(dir + list[j]); ref = File.getName(dir + list[j]); imageCalculator("Difference create 32-bit", currentFile, ref); //run("Image Calculator...", "image1="+currentFile+" operation=Difference image2="+ref+" create 32-bit"); run("Square"); showStatus("Summing pixel values..."); width = getWidth(); height = getHeight(); for (y=0; y<height; y++) { if (y%20==0) showProgress(y, height); for (x=0; x<width; x++) { value = getPixel(x,y); sum = sum + value; } } close(); close(); if (sum<large){ tick = ref; large = sum; } sum = 0; } print(tick+","+round(xcentroid+xcenter)+","+round(ycentroid+ycenter)); close(); xcenter = xcorner + xcenter - 15; ycenter = ycorner + ycenter - 5; } 26 Appendix D: Preliminary Data Analysis of Pressure-Drive Flow Velocity vs. Y for Z = 0 microns Velocity vs. Y for Z = 14.4 microns 100 250 90 200 70 Velocity (um/s) Velocity (um/s) 80 60 50 40 30 150 100 50 20 0 10 0 R2 = 0.3612 10 20 30 40 0 10 20 30 40 50 60 R2 = 0.9972 0 50 -50 60 Y (microns) Y (microns) Velocity vs. Y for Z = 19.2 microns Velocity vs. Y for Z = 9.6 microns 250 250 200 Velocity (um/s) 150 100 150 100 50 50 0 0 R2 = 0.9864 10 20 30 -50 0 10 20 30 40 50 60 Y (microns) Y (microns) Velocity vs. Position 60 50 40 30 20 10 0 0 10 40 50 R2 = 0.9594 0 Z (microns) Velocity (um/s) 200 30 20 40 50 60 -10 Y (microns) (The area of each circle corresponds to the velocity of the bead at that position.) 27 60 28